A machine learning (ML) model is part of the match model. You can train an ML model to predict whether record pairs are a match or not. A trained ML model can analyze data attributes and make match predictions based on its learning.
Before you train an ML model, you need to identify the training data set that is a true representation of your organization's data. The results produced by the ML model are only as good as the training data set the model is trained on.
During the training process, users label record pairs that are presented to them in batches. The labels indicate whether the record pairs are a match or not. The labeled record pairs are the examples that the model learns from to predict whether record pairs are a match.
To train an ML model, use a custom match model that suits your business needs. You create a custom model from a copy of a predefined match model or from scratch.
After you publish a match model, you can define match and merge jobs, and run the jobs. For information about defining and running match and merge jobs, see Define Jobs.